We have been constantly facing an issue that brings Flink down, that is, upstream bad data breaks the CustomDeserializationSchema step, and then the whole pipeline went down. Is there a good way that can prevent this from happening, and only log / report metric on such cases?